For years now, many AI industry watchers have looked at the quickly growing capabilities of new AI models and mused about exponential performance increases continuing well into the future. Recently, though, some of that AI "scaling law" optimism has been replaced by fears that we may already be hitting a plateau in the capabilities of large language models trained with standard methods.

The verdict is in: OpenAI's newest and most capable traditional AI model, GPT-4.5, is big, expensive, and slow, providing marginally better performance than GPT-4o at 30x the cost for input and 15x the cost for output. The new model seems to prove that longstanding rumors of diminishing returns in training unsupervised-learning LLMs were correct and that the so-called "scaling laws" cited by many for years have possibly met their natural end.

TL;DR: I chose to make using AI a manual action, because I felt the slow loss of competence over time when I relied on it, and I recommend everyone to be cautious with making AI a key part of their workflow.

"We're in the very early days looking at this problem from an ecosystem level," Larson told The Register. "It's difficult, and likely impossible, to quantify how many attempted installs are happening because of LLM hallucinations without more transparency from LLM providers. Users of LLM generated code, packages, and information should be double-checking LLM outputs against reality before putting any of that information into operation, otherwise there can be real-world consequences."

[...]

"Even worse, when you Google one of these slop-squatted package names, you’ll often get an AI-generated summary from Google itself confidently praising the package, saying it’s useful, stable, well-maintained. But it’s just parroting the package’s own README, no skepticism, no context. To a developer in a rush, it gives a false sense of legitimacy.

IA et vie privée : selon les spécialistes de la sécurité, tout n'est pas si rose et malgré le grand mixage qu'est la phase d'entraînement du modèle de données, il est tout à fait possible que des données privées soient involontairement préservées, et donc publiquement accessibles dans le modèle final.

I think most privacy experts would agree with this post so far. There are divergences of opinion when you start asking "do the benefits of AI outweigh the risks". If you ask me, the benefits are extremely over-hyped, while the harms (including, but not limited to, privacy risks) are very tangible and costly. But other privacy experts I respect are more bullish on the potentials of this technology, so I don't think there's a consensus there.

AI companies, however, do not want to carefully weigh benefits against risks. They want to sell you more AI, so they have a strong incentive to downplay the risks, and no ethical qualms doing so. So all these facts about privacy and AI… they're pretty inconvenient. AI salespeople would like it a lot if everyone — especially regulators — stayed blissfully unaware of these.

Aucune technologie n’est neutre ni inéluctable. Chacune se déploie dans un certain contexte économique et politique qui oriente les choix. Cela a toujours été le cas pour le numérique, depuis le début. L’extrême concentration d’acteurs et de moyens qui préside au déploiement des IAs génératives devrait aider à prendre conscience de cet état de fait. L’annonce récente de 500 milliards de dollars à consacrer au sujet donne la (dé)mesure de la chose. Je ne détaillerai pas les courants politiques et philosophiques qui circulent parmi les promoteurs des IAs. Certains acteurs affirment croire à l’avénement des IAs générales, comme résultat inéluctable de l’accumulation de moyens et de ressources. Que l’on fasse miroiter ces IAs capables de sauver le monde, ou qu’au contraire on annonce l’apocalypse, leur prise de pouvoir et la fin de l’humanité, on participe à détourner l’attention des dégâts déjà bien présents ici et maintenant.

In a recent earnings call Sundar Pichai claimed that at Google now 25% of Code is AI generated (“and then reviewed and accepted by engineers”). In the AI boosterism parts of the web (so basically X and LinkedIn) this number was celebrated: Even Google does AI code generation. So if your whole startup is just ChatGPT in a trenchcoat, you’re basically at the industry standard, right?

Let’s not be cynical here and point at Google’s not exactly stellar recent track record when it comes to great products and software, but let’s ask us where that number comes from and what it means.

I went to the UX Brighton conference yesterday.

The quality of the presentations was really good this year, probably the best yet. Usually there are one or two stand-out speakers (like Tom Kerwin last year), but this year, the standard felt very high to me.

But…

The theme of the conference was UX and “AI”, and I’ve never been more disappointed by what wasn’t said at a conference.

As OpenAI and Meta introduce LLM-driven searchbots, I'd like to once again remind people that neither LLMs nor chatbots are good technology for information access. [...]

If someone uses an LLM as a replacement for search, and the output they get is correct, this is just by chance. Furthermore, a system that is right 95% of the time is arguably more dangerous tthan one that is right 50% of the time. People will be more likely to trust the output, and likely less able to fact check the 5%.

But even if the chatbots on offer were built around something other than LLMs, something that could reliably get the right answer, they'd still be a terrible technology for information access.